Automatic Generation of Indicative Summaries for Multiple Documents

Title: Automatic Generation of Indicative Summaries for Multiple Documents

Research Question: How can natural language generation be used to create indicative summaries for multiple documents?

Methodology: The researchers analyzed the task of creating indicative summaries, which help users decide if they want to read a document. They used a natural language generation model to create indicative multidocument summaries. This model was based on high-level document features, such as the distribution of topics and media types.

Results: The researchers found that their model could effectively create indicative summaries that highlighted the differences between documents and were relevant to user queries. They implemented this model in their C ENTRIFUSER summarization system and generated a sample indicative multidocument query-based summary.

Implications: This research suggests that natural language generation can be used to create indicative summaries for multiple documents. This can help users make better decisions about which documents to read. The researchers' model provides a framework for creating such summaries and could be further developed and applied in other summarization systems.

Link to Article: https://arxiv.org/abs/0107019v2 Authors: arXiv ID: 0107019v2